Research On The Key Technology Of Large Field Of View And Ultra-High Resolution Camera

Perception of the environment is the main approach to get information. Above 80% of the total accessed information is visual information. Image sensors have brought great impact on research, industrial, daily life and other areas. Image sensor technology continues to progress with widely and further used. The image instrument with large field of view (FOV), high resolution and high frame rate is the goal of scientific researchers and engineers. Based on advanced integrated circuit chip technology, we have proposed a flat geometry multi-camera solution and established an experiment instrument after comparing several of large FOV and high resolution schemes.The main research contents of this dissertation are summarized as follows:1. First, three aspects of the instrument have been introduced which are mechanical structures, electrical drives and data acquisition of image sensors, and image data transmission protocols. We have studied on the relationship between spherical FOV and rectangle FOV and obtained a segmentation method for the spherical FOV taking rectangle FOV as unit. The details of the optical axis of sub-cameras point to and the segmentation result of spherical FOV have been presented. Considering the key of FOV segmentation is the optical axis direction and the size of rectangle FOV unit and keeping both of these two constraints unchanged, the image sensors can be shifted within local space. So we obtain a flat camera array geometry structure which is similar to Fresnel lens. This structure has the advantage of flexible, compact and convenient for electrical connection. We have studied on the method of data acquisition of multiple image sensors based on field-programmable-gate-array (FPGA) device. The functions including exposure control of the sensors, data acquisition and high-speed digital image transmission protocol are finished on a single chip. Consequently, the system is more compact and more stable. And the performances of the system have tested.2. An ideal image system should be a linear space-invariant system with aberration-free, noise-free. An available image system limited by various factors unable to reach this ideal state. The photometric response characteristics of the large FOV camera system were calibrated. We can obtain numerical relationship between the pixel value and illumination entranced at pupil by the calibration process. The large FOV imaging system consists of multiple sensors face a problem which is same object point may generate different pixel value on multiple cameras caused by lens, sensitivity of sensors and electronic components differences. In order to quantify the impact of such difference on the final panoramic images, each sub-camera photometric response characteristics has been calibrated. Parameters of calibration include linear and nonlinear response range, dark current and vignetting cause by lens and mechanical structures. We hve present the corrected example image and panorama image. Range of luminance value is very large in the case of imaging with large FOV. Dark area and extream bright area are unable to get a reasonable pixel value when using wide-angle lens with a single sensor for wide-field imaging for the dynamic limit of the sensor. The limitations can be relieved through each sensor covering partial field of view and then composed a large field of view as indicate in our research. The photometric characteristics of our large FOV image system have been comprehensive analyzed, providing a good support for its further application based on the multi-camera scheme.3. In order to get the precision internal and external parameters, each sub-camera has been calibrated. The distortions of images output by cameras are corrected at first based on the calibrated internal parameter results. We have studied the method of image stitch method and panorama display. Combining the internal and external parameters, we have developed an absolute registration method for image stitching in our large FOV image system. Therefore, we can avoid registering images one by one. The difficult of image stitch caused by moving objects in the scene is missing using multiple sensor exposure at the same time. Multi-band blending method has been used to blend registered images. A panorama with various project surfaces can be obtained through our image system. The compact flat geometry multi-camera large FOV image system will act an important role in virtual reality,3D reconstruction, object tracking and much area.